Prediction Market Arbitrage

Cross-platform arbitrage scanner for Kalshi and Polymarket, plus paper trading system.

Status: Scanner Running + Paper Trading Active 📊

Location: ~/clawd/prediction-arb/

Components

1. Arb Scanner (Hourly Cron)

Scans Kalshi and Polymarket for pricing discrepancies.

Reality check (learned the hard way):

• Pre-game arbs basically don't exist (1-2¢ differences)
• Live game arbs exist but close in seconds — too fast for browser-based execution
• Long-dated arbs are real (4-8% on NBA championship) but lock capital for months
• Short-term efficiently priced; long-term = opportunity but patience required

2. Paper Trading Dashboard

Web UI tracking 10 paper positions with live Polymarket prices.

URL: Local dashboard (needs hosting decision)

Active Positions:

• 2 NBA ROY awards (resolve Feb 6)
• 6 Super Bowl halftime positions (resolve Feb 9)
• 2 economic indicators

Strategy: Bayesian bias exploitation — buy underpriced outcomes, apply probability updates as news drops.

Key Learnings

Insight	Date	Impact
Pre-game arbs don't exist	Feb 2	Pivoted to bias exploitation
Live arbs close in seconds	Jan 30	Browser too slow, need API execution
Long-dated arbs are real	Jan 30	4-8% guaranteed but capital lock
Polymarket US = sports only	Jan 31	CFTC restrictions limit entertainment markets
Paper trade fast feedback	Feb 2	Short-term positions (≤30 days) validate strategy

Cron Jobs

• arb-scanner — Every hour, scans for opportunities
• live-arb-monitor — Every 10min during NBA hours (experimental)

Files

• scanner/ — Market scraping and comparison logic
• dashboard/ — Paper trading web UI
• data/ — Historical price data

Next Steps

• Review ROY results (Feb 6)
• Apply Bayesian updates to Super Bowl positions as news drops
• Decide on dashboard hosting (Tailscale vs Cloudflare tunnel)
• If paper trading profitable, consider real capital allocation

---

Created: 2026-01-28 | Last updated: 2026-02-02